基于累积和算法的动态变形预警模型研究

Research on Warning Model for Dynamic Deformation Based on CUSUM

利用GPS对构筑物进行动态变形监测,获取构筑物的动态变形序列,通过卡尔曼滤波处理变形时间序列,从而削弱相关噪声的影响,提取真实的变形信号,并通过FFT处理变形信号,验证其滤波效果。在分析累积和算法的前提下,构建构筑物预警模型,分析了平均运行长度及预警阈值等相关参数的选取原则。经实测数据分析,模型能有效检测到变形发生时刻,该方法能够提高GPS进行动态变形预警的可靠性,特别是为微小变形提供一定研究指导意义。

By using GPS to monitor dynamic deformation of structures,dynamic deformation series were acquired.And the deformation time series are processed with Kalman to reduce the influance of correlated noise and extract the real deformation signal.Then the deformation is processed with FFT to verify the filter effect.On the premise of analyzing CUSUM algorithm the structure warning model was built,and the choosing rule of average run length（ARL） and warning threshold are presented.The model is successfully applied to the warning with the field collected data,and it can detect the time of deformation effectively,the method can improve the reliability of deformation warning by using GPS,especially for researching the small deformation.